{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/mirrorgan-learning-text-to-image-generation","title":"MirrorGAN: Learning Text-to-image Generation by Redescription","arxiv_id":"1903.05854","date":"2019-03-14","proceeding":"CVPR 2019 6","authors":["Tingting Qiao","Jing Zhang","Duanqing Xu","DaCheng Tao"],"abstract":"Generating an image from a given text description has two goals: visual\nrealism and semantic consistency. Although significant progress has been made\nin generating high-quality and visually realistic images using generative\nadversarial networks, guaranteeing semantic consistency between the text\ndescription and visual content remains very challenging. In this paper, we\naddress this problem by proposing a novel global-local attentive and\nsemantic-preserving text-to-image-to-text framework called MirrorGAN. MirrorGAN\nexploits the idea of learning text-to-image generation by redescription and\nconsists of three modules: a semantic text embedding module (STEM), a\nglobal-local collaborative attentive module for cascaded image generation\n(GLAM), and a semantic text regeneration and alignment module (STREAM). STEM\ngenerates word- and sentence-level embeddings. GLAM has a cascaded architecture\nfor generating target images from coarse to fine scales, leveraging both local\nword attention and global sentence attention to progressively enhance the\ndiversity and semantic consistency of the generated images. STREAM seeks to\nregenerate the text description from the generated image, which semantically\naligns with the given text description. Thorough experiments on two public\nbenchmark datasets demonstrate the superiority of MirrorGAN over other\nrepresentative state-of-the-art methods.","url_abs":"http://arxiv.org/abs/1903.05854v1","url_pdf":"http://arxiv.org/pdf/1903.05854v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"mirrorgan-learning-text-to-image-generation","repo_url":"https://github.com/Kyfafyd/MirrorGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"mirrorgan-learning-text-to-image-generation","repo_url":"https://github.com/komiya-m/MirrorGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"image-generation","task_name":"Image Generation"},{"task_slug":"image-to-text","task_name":"Image to text"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"text-to-image-generation-1","task_name":"Text to Image Generation"},{"task_slug":"text-to-image-generation","task_name":"Text-to-Image Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/text-to-image-generation-on-cub","task":"Text-to-Image Generation","dataset":"CUB","model":"MirrorGAN","rank_in_archive_order":18,"of":20,"metrics":{"Inception score":"4.56"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1903.05854","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}